Basser Seminar Series

Making Machine Learning Useable with Transparent Approach and Measurable Decision Making

Speaker: Dr Fang Chen

Time: Wednesday 18 September, 4:00-5:00pm
Refreshments will be available from 3:30pm

Location: The University of Sydney, School of IT Building, Lecture Theatre (Room 123), Level 1

Add seminar to my diary


Machine learning (ML) techniques are often found difficult to be effectively applied in practice because of their complexities, such as complex parameter settings and mathematical models. Therefore, making ML useable is emerging as one of active research fields recently. For example, an ML algorithm is often regarded as a “black-box”, where the user defines parameters and input data for the “black-box” and gets output from its execution. This “black-box” approach makes it difficult for users to understand the complicated ML models. As a result, the user is uncertain about the usefulness of ML results and this affects the effectiveness of ML methods. Furthermore, in the most cases, making decision is the ultimate goal of an ML-based data analysis process. And when an ML approach is used to infer a model from input data, the quality of the model should be judged from the point of view of how good are the decisions one makes based on this model. Therefore, there is a close connection between ML and decision theory. This connection becomes one of significant parts in making ML useable.

Our current research focuses two aspects: 1) Our research makes a “black-box” ML process transparent by presenting real-time internal status update of the ML process to users explicitly. A user study was performed to investigate the impact of revealing internal status update of ML to users on the easiness of understanding the data analysis process, meaningfulness of real-time status update, and convincingness of the ML results. The study showed that revealing of the internal states of ML process can help to improve easiness of understanding the data analysis process, make real-time status update more meaningful, and make ML results more convincing. 2) The connection between ML and decision theory is set up with a measurable decision making process by Galvanic Skin Response (GSR). The measurable decision making based on ML results was studied to investigate the impact of ML results on decision making process. It showed that the number of ML-based decision factors, the value of an ML-based decision factor as well as different ML-based decision factors affect the easiness of decision making process differently. The results can help to provide guidelines for ML-based applications in the effective use of ML results in decision making.

Speaker's biography

Dr Fang Chen was employed with Beijing Jiaotong University in China for various positions, including as the Dean of the Faculty of Electronic and Information Engineering in 1997. She began her career in industry in 1999 at Intel China Research Centre. Subsequently she joined Motorola, at which she served different roles, including as founding manager of the Speech and Language Generation Lab, business relationship manager of Motorola China Research Centre, and Patent and Publication Committees Chair of the Motorola Australian Research Centre. She joined NICTA in 2004 and is currently the Research Group Manager at the ATP Laboratory in the Machine Learning Research Group.

Dr Chen's main research interest is behaviour analytics, including human-machine interaction, cognitive load modelling, signal processing, pattern recognition, machine learning, and the prediction and evaluation of human performance and system performance. She has been actively involved in a multitude of research and development activities, including actively pursuing a start-up company and technology licensing. She has led a range of projects, including mobile applications (using speech, handwriting, natural language as well as providing location based services), multimodal human mental status monitoring (such as cognitive load, stress and human trust), data driven traffic modelling and decision support systems, and infrastructure asset management (such as water pipe, bridge). She has more than 130 refereed publications and has filed 30 patents in Australia, the US, Europe, Canada, China, Japan, Korea and Mexico.

Dr Chen is a Conjoint Professor with the University of New South Wales, and an adjunct professor with Beijing Jiaotong University.